[rsync@b] Re: Dirvish, --link-dest and permissions

jw schultz jw at pegasys.ws
Sat Mar 15 17:28:10 EST 2003

On Sat, Mar 15, 2003 at 12:38:49AM -0500, Bert wrote:
> >The network traffic in this case is the blocksums and file
> >block copy instructions.  This is the same traffic that you
> >would get if you updated the timestamps.
> Hmmm. Here is the stats output at the end of one of these 
> all-permissions-have-changed rsync sessions:
> (this indicates to me that rsync used no local data and sent everything 
> over the wire)
> 	Number of files: 40817
> 	Number of files transferred: 37757
> 	Total file size: 4031988583 bytes
> 	Total transferred file size: 4031988573 bytes
> 	Literal data: 4031988573 bytes
> 	Matched data: 0 bytes
> 	File list size: 1263669
> 	Total bytes written: 604198
> 	Total bytes read: 4035210974
> 	wrote 604198 bytes  read 4035210974 bytes  347331.23 bytes/sec
> 	total size is 4031988583  speedup is 1.00
> Also, here is the rsync command that dirvish invoked:
> (perhaps interesting is the whole-file option?)
> ACTION: rsync -v --stats -a -H --delete --delete-excluded --numeric-ids 
> --exclude-from - -W --link-dest /usr/local/data/bac
> kups-dirvish/pcdirs-home/20030310-12:55/tree 
> localhost:/usr/local/data/pc-homedirs/home/ 
> /usr/local/data/backups-dirvish/pc
> dirs-home/20030314-19:50/tree | sed -e '/\/$/d' -e '/ [-=]> /d' >> 
> /usr/local/data/backups-dirvish/pcdirs-home/20030314-19:
> 50/log
> Finally, here is an indication that the 20030310 and 20030314 backups were 
> similar:
> 	bash-2.05b# du -s 2003031[04]*
> 	3967478 20030310-12:55
> 	4083962 20030314-19:50
> 	(diff on the first 10,000 filenames showed only 4 changes)
> Is this what you expected to see and what you meant by "the same traffic 
> that you would get if you updated the timestamps"?  I honestly don't know 
> what is expected if the timestamps update, though I vaguely remember lots 
> of CPU (calc checksums) but not nearly as much network traffic as this when 
> daylight savings kicked in & the FAT timestamps got confused.

With those options a timestamp change would have had the
same effect.  Using the whole-file option will disable the
rsync algorithm so, yes, you will see no use of the local

Even if a copy were made when only the meta-data changed i
wouldn't do so if -W is applied.  With a fast network local
copy is often slower than a network copy due to the disk

If you are using dirvish set the client field to the output
of hostname, not localhost, and it will do a local copy.

	J.W. Schultz            Pegasystems Technologies
	email address:		jw at pegasys.ws

		Remember Cernan and Schmitt

More information about the rsync mailing list